Crate pdf_engine 

Unified PDF rendering engine.

pdf-engine is the main public-facing API for reading and rendering PDF documents. It wraps the lower-level pdf-syntax / pdf-interpret / pdf-render stack and exposes a single PdfDocument handle for all common operations: page rendering, text extraction, thumbnails, metadata, bookmarks, and full-text search.

Quick Start

use std::sync::Arc;
use pdf_engine::{PdfDocument, RenderOptions};

// Load from bytes (accepts Arc<Vec<u8>>, Vec<u8>, or any Into<PdfData>).
let data = Arc::new(std::fs::read("invoice.pdf").unwrap());
let doc = PdfDocument::open(data).unwrap();

println!("{} pages — {:?}", doc.page_count(), doc.info().title);

// Render page 0 at 150 DPI → raw RGBA pixel data.
let opts = RenderOptions { dpi: 150.0, ..Default::default() };
let rendered = doc.render_page(0, &opts).unwrap();
println!("{}×{} px", rendered.width, rendered.height);

// Plain-text extraction.
let text = doc.extract_text(0).unwrap();
println!("{text}");

// Structured text with per-span positions.
for block in doc.extract_text_blocks(0).unwrap() {
    for span in &block.spans {
        println!("  [{:.0}, {:.0}] {}", span.x, span.y, span.text);
    }
}

// Full-text search — returns 0-based page indices.
let hits = doc.search_text("total");
println!("'total' found on {} page(s)", hits.len());

Key Types

Type	Description
BatchConfig / BatchResult	Worker-pool processing for many PDFs
PdfDocument	Main document handle
RenderConfig / RenderOptions	DPI, color mode, background colour, optional forced width/height
RenderedPage	RGBA or CMYK pixel data (row-major, 4 bytes per pixel)
PageGeometry	MediaBox, CropBox, TrimBox, BleedBox, rotation
PageBox	A rectangle in PDF user-space points
DocumentInfo	Title, author, subject, creator, producer
TextBlock / TextSpan	Structured text with position and font size
Block / BlockKind / ExtractOptions	Paragraphs, headings and list items, and which pages of them come back
Word	One word with its box, page, block and line
to_markdown / to_json	Those blocks as markdown for a reader, or as JSON for a chunker
BookmarkItem	Outline node — title, target page, nested children
ThumbnailOptions	Max-dimension constraint for thumbnail rendering

Re-exports

pub use batch::process_batch;

pub use batch::process_batch_with_page_progress;

pub use batch::BatchConfig;

pub use batch::BatchResult;

pub use batch::ErrorStrategy;

pub use batch::PdfBatch;

pub use blocks::Block;

pub use blocks::BlockKind;

pub use blocks::ClipMargins;

pub use blocks::ExtractOptions;

pub use blocks::OcrMode;

pub use blocks::PageSeparators;

pub use blocks::Quality;

pub use blocks::TableRenderMode;

pub use blocks::TableStrategy;

pub use blocks::TatrOptions;

pub use color::preserve_device_cmyk;

pub use docling::to_docling_json;

pub use docling::DoclingPage;

pub use document::join_page_texts;

pub use document::BookmarkItem;

pub use document::DocumentInfo;

pub use document::PdfDocument;

pub use error::EngineError;

pub use error::OptionsError;

pub use error::Result;

pub use geometry::PageBox;

pub use geometry::PageGeometry;

pub use geometry::PageRotation;

pub use json::to_json;

pub use limits::LimitError;

pub use limits::ProcessingLimits;

pub use markdown::to_markdown;

pub use markdown::to_markdown_with_code;

pub use markdown::to_markdown_with_options;

pub use markdown::MarkdownOptions;

pub use ocr::OcrBackend;

pub use ocr::OcrError;

pub use ocr::OcrResult;

pub use ocr::OcrWord;

pub use ocr_need::page_needs_ocr;

pub use ocr_need::OcrNeed;

pub use render::ColorMode;

pub use render::PixelFormat;

pub use render::RenderConfig;

pub use render::RenderOptions;

pub use render::RenderedPage;

pub use text::BoundsSource;

pub use text::FontMetrics;

pub use text::GeometryMode;

pub use text::LayoutBlock;

pub use text::TextBlock;

pub use text::TextSpan;

pub use text::WidthSource;

pub use text_info::TextSpanInfo;

pub use thumbnail::ThumbnailOptions;

pub use words::to_json_with_words;

pub use words::to_words_json_lines;

pub use words::Word;

pub use ocr::best_available_backend;

pub use ocr::ocr_page_default;

pub use ocr::AwsTextractBackend;

pub use ocr::AzureDocIntelBackend;

pub use ocr::GoogleVisionBackend;

pub use ocr::MistralOcrBackend;

pub use ocr::OcrsBackend;

pub use ocr::PaddleOnnxBackend;

Modules

api

The ideal top-level API facade for the PDFluent SDK.

api_error

Public engine error type and stable error-code mapping. See [api_error::PdfEngineError].

balanced

The balanced tier’s page assembly: detector boxes plus our own text (#641). Off by default, beside layout. The balanced tier’s page assembly: detector boxes plus our own text, page in, markdown out (#641).

batch

Batch helpers for processing many PDFs with a bounded worker pool.

blocks

Paragraphs and the blocks::Block model the structured serializers consume (#567). Paragraphs out of the lines a page draws (#567).

color

Color helpers for render output conversion.

docling

The block model as a Docling DoclingDocument (#629).

document

Unified document facade — multi-page rendering, text extraction, metadata, bookmarks, and thumbnails.

error

Error types for the rendering engine.

geometry

Page geometry: boxes (MediaBox, CropBox, TrimBox, BleedBox, ArtBox), rotation, and DPI-based pixel conversions.

json

The block model as JSON: every block with its kind, page, box, text, spans and reading-order index, for a caller that chunks the document itself (#572). The block model as JSON, for callers that chunk it themselves (#572).

layout

Layout detection on the shared ONNX Runtime layer (#641). Off by default. Layout detection on the shared ONNX Runtime layer (#641).

limits

Resource limits for PDF processing.

markdown

The block model as markdown: headings, lists, tables and a marker where the page changed (#572). The block model as markdown (#568, #569, #570, #571).

markdown_images

Figures in the block model: listing, naming, pixels, and references (#658). Figures in the block model: listing, naming, pixels, and references.

model

The model tiers’ shared surface: specs, failures and install paths (#641). Always compiled; only the session half needs the ort feature. The model tiers’ shared surface: specs, failures and install paths (#641).

ocr

OCR backend trait and implementations.

ocr_need

Scan detection: how much OCR a page needs, before any recognizer runs. Scan detection: how much OCR a page needs, before any recognizer runs.

render

Page rendering with z-order compositing.

text

Text extraction via a custom Device implementation.

text_info

Canonical, serializable wire form of an extracted text span.

thumbnail

Thumbnail generation options.

vision

Optional vision layer: raster plus ONNX runner (#636). Off by default. Optional vision layer: page raster plus ONNX runner (#636).

words

Word-level boxes for citations, highlighting and grounding.

words_join

The rule that decides whether two adjacent pieces read as one word.

xfa

XFA feature surface for pdf-engine.

Enums

RasterQuality

Rasterization precision / speed trade-off for the vello_cpu pipeline.